Large-scale yeast proteome analysis using multi dimensional protein identification technology (MuDPIT
1. Large-scale analysis of the yeast proteome by multi dimensional protein identification technology Michael P. Washburn, Dirk Wolters and John R. Yates Nature Biotechnology Volume 19 Pages 242-247 PRIYANK JAIN NISHANT CHORADIA DIVYA KALRA 09 OCTOBER’08
2. THE DIFFERENT MINING APPROCHES 2D-PAGE & MS MuDPIT Well-suited to simpler samples where the goal is to characterize major system components Advantages offers high throughput takes advantage of powerful protein-separation methodology Unable to identify proteins with extremes in pI and molecular weight low – abundance proteins membrane associated or bound proteins For exhaustive mining of both high- and low-abundance proteins in complex mixtures, MuDPIT is the most effective approach Advantages generates the most reliable protein identification because it is based on MS-MS spectra, which directly indicates peptide sequences Technically challenging and still rapidly evolving
3. WHY MuDPIT? Previous proteomic analyses of the S.cerevisiae yielded 279 proteins in a single study using 2D-PAGE coupled to MS Wide variety of systems coupling multidimensional chromatography to MS have been used but none identified > 200 proteins from any sample A fully automated high-throughput method was needed that combined resolution and identification removing all sample-handling steps once the sample has been loaded. A fully online 2D LC/MS/MS system like MuDPIT fulfills both of these requirements MuDPIT – resolution of peptides and the generation of tandem MS occur simultaneously MuDPIT may specifically be applied to integral membrane proteins to obtain detailed biochemical information on this unwidely class of proteins
4. EXPERIMENT S.cerevisiae was grown till mid-log phase, lysed and three different fractions were generated for analysis – soluble fraction, lightly washed insoluble fraction and heavily washed insoluble fraction Digestion of the soluble fraction was done using Endoproteinase Lys-C and trypsin and a complex peptide mixture was prepared for amino acid analysis on each sample Digestion of the insoluble fractions was done using formic acid and CNBr and a complex peptide mixture was prepared for amino acid analysis on each sample MuDPIT analysis was done on each sample SEQUEST algorithm was run on each of the three data sets against the yeast-orfs.fasta database
6. RESULTS MuDPIT method is reproducible on the levels of both the chromatography and the final protein list The results are from the runs of three separate fractions After combining the MS/MS data generated 5,540 peptides were assigned to the MS spectra leading to the identification of 1,484 proteins from the S.cerevisiae proteome. Proteins identified in the AUTOQUEST were further analyzed using MIPS S.cerevisiae catalog The analysis showed that the results provide a representative sampling of the yeast proteome The results also proved that the MuDPIT method was largely unbiased
13. DISCUSSION Method used in this study provides a large-scale and global view of S.cerevisiae proteome Determined the proteins in a largely unbiased manner The sensitivity level across class of proteins listed ranged from 13% of the predicted proteins identified with pIs < 4.3 and MWs <10kDa to 43% of the predicted proteins identified with pIs >11. Method had slight biased against proteins with pIs < 4.3 and MWs < 10kDa Decreased sensitivity in these class of proteins – likely because of lack of tryptic peptides in the final mixture
14. CONCLUSION This work was a major step towards high-throughput methods because it was able to detect the low-abundance proteins, peripheral and integral membrane proteins. When emerging quantitative proteomic methods are combined with MuDPIT, true large-scale analysis of protein expression changes will be possible. Combination of MuDPIT with quantitative methods will allow for the integration of mRNA and protein expression levels needed to fully understand gene networks.
Hinweis der Redaktion
As CNBr cleaves at methionine residues and leaves either homo serine or homo serine lactone.Formic acid partially solubilizes the membrane portions of the cell in the heavily and lightly washed insoluble fractions.CNBr cleaves off the soluble portions of the integral membrane proteins as large domains that were subjected to additional proteolysisMS/MS data resulting from the 2 samples had to be independently analysed twice with SEQUESTFor each run differential search modification was usedThree charge states identified with SEQUEST +1, +2,+3For +1 peptideds, they were accepted if they were fully tryptic and had cross correlation (Xcross)of 1.9For +2 peptides, they were accepted if they were fully tryptic and has X Cross values of 2.2 and 3.0For +3 peptides, thery were accepted if they were fully or partially tryptic and had a XCorr>3.75
Complex peptide mixtures from different fractions of a S. cerevisiae whole-cell lysate were loaded separately onto a biphasic microcapillary column Microcapillary column – packed with reverse phase material and a strong cation exchange materialColumn was inserted into the instrumental setupHPLC coupled to LCQ ion trap mass spectrometer equipped with a nano-LC electrospray ionisation sourcePeptides directly eluted into the tandem mass spectrometer because a voltage (kV) supply is directly interfaced with the microcapillary columnPeptides were first displaced from the SCX to the RP by a salt gradient and eluted off the RP into the MS/MSIn an iterative process, the microcolumn was re-equilibrated and an additional salt step of higher concentration displaced peptides from the SCX to the RPPeptides were again eluted by an RP gradient into the MS/MS, and the process was repeatedFully automated 15-step chromatography run was carried out on each sampleThe tandem mass spectra generated were correlated to theoretical mass spectra generated from protein or DNA databases by the SEQUEST algorithm
Chromatography reproducibility is the identification of the same peptide at the same point in the chromatography in two or more separate analysis.
a) Subcellular localizations obtained from the S. cerevisiaesubcellular localization catalog at the Munich Information Center for Protein Sequences website.b) Proteins identified in individual runs were analyzed for their subcellular localization. The subcellular localization of many of the proteins detected and identified is unknown. Therefore, not all of the proteins detected and identified are represented in this table.The subcellular localization catalogs from MIPS allowed us to determine the similarities and differences among the three fractions (Table 1). Eventhough in several cases the overall numbers of proteins identified from a cellular compartment appear similar between any two samples, unique identifications were found in every sample. For example, the majority of the unique hits from the soluble fraction were proteins localized to the cytoplasm and the nuclei of S. cerevisiaeincluding the transcription factor SNF5 .The two insoluble fractions provided greater detections and identifications of organelle proteins . The heavily washed insoluble fraction had more hits than any other sample localized to the nucleus, mitochondria, endoplasmic reticulum, plasma membrane, and Golgi. There were unique hits found in both the heavily washed insoluble fraction and partially washed insoluble fraction. Of the major protein classes rarely seen on 2D-PAGE, we detected and identified 32 protein kinases including the MAP kinase signal transduction pathway kinases. Furthermore, they detected and identified 45 transcription factors.
Figure 2. Codon adaptation index (CAI) distribution of the identified S. cerevisiaeproteome and the predicted S. cerevisiae genome. (A) CAI distribution of the proteins predicted in the S. cerevisiae genome. (B) Compare this to the distribution of the proteins identified in this study over CAI ranges. In both cases, the largest protein region is found between the CAI range of 0.11 and 0.2. (C) The average number of peptides identified for each protein in a particular CAI range was determined and plotted against CAI rangesOf the 6,216 open reading frames in the yeast genome, 83% have CAI values between 0 and 0.20, that is, are predicted to be present at low levels. Previous proteomics studies in yeast have identified few proteins with CAIs <0.2 (refs. 4,5,32). Efforts are underway to overcome these shortcomings of 2D-PAGE, but recent evidence suggests that 2D-PAGE alone is incapable of detecting low-abundance proteins. Any large-scale proteomic analysis of S. cerevisiae must identify proteins in this CAI range. As seen in Figure 2B, the data from our study yield a representativesample of the yeast proteome with 791 or 53.3% of the proteins identified having a CAI of <0.2. A total of 1,347 peptides were detected from the 791 proteins identified with a CAI of <0.2, an average of 1.7 peptides per protein. The number of peptides per protein increases with increasing CAI (Fig. 2C). Because CAI is considered a predictor of protein abundance, the most abundant proteins are the easiest to detect in any sample resulting in more peptide identifications from abundant proteins than low-abundance proteins.
Because a peptide mixture is generated before the chromatography, the method should be independent of pI and MW of proteins. In two of the studies for which MW and pI were reported for the proteins identified, no protein with a MW >180 kDa or pI >10 was detected and identified.Proteins with both acidic and basic pIs are represented in our data set. Twelve proteins with pIs <4.3 were identified, with the lowest being RPP1A (YDL081C), which has a pI of 3.82 (data not shown). Twenty-nine proteins with pIs >11 were identified, with the most basic protein identified being RPL39 (YJL189W), which has a pI of 12.55 (data not shown). In addition, proteins with MWs <10,000 and >190,000 Da are represented. For example, 24 out of 77 possible proteins with a MW in excess of 190 kDa were identified, the largest being YLR106C (CAI = 0.17) with a MW of 558,942 Da, from which four unique peptides were identified.
By analyzing our data set against the peripheral membrane proteins contained in the Yeast Proteome Database, we detected and identified 72 out of 231 possible peripheral membrane proteins. We uniquely detected 23 in the heavily washed insoluble fraction and 14 from the lightly washed insoluble fraction (data not shown).At the MIPS website24, the entire yeast genome has been analyzed for loci with predicted transmembrane (Tm) domains from 1 to 20. Using these , 697 proteins from the S. cerevisiae genome have three or more predicted Tm domains, of which we identified 131 or 19% of the total(Table 2). Of these 131 proteins, 44 were identified only in the heavily washed insoluble fraction, and 33 were identified only in the lightly washed insoluble fraction. Several of these proteins have low predicted abundances based on their CAI.
Two unique peptides were detected for the poorly characterized protein YCR017c (CAI =0.16), which has 15 predicted Tm domains (Table 3)24.The peptides detected and identified from each predicted integral membrane protein rarely covered part of or all of a predicted Tm domain (Table 3). Of the 70 peptides identified from 26 proteins with 10 or more predicted Tm domains, 4 peptides partially covered predicted Tm domains (FKS1, ALG7, and YGR125w) and 4 peptides completely covered predicted Tm domains (ALG7, ITR1, PMA1, and PMA2) (Table 3). Furthermore, 43 of the 70 peptides listed in Table 3 mapped to the largest soluble domain of the respective protein.These patterns persisted with the identifications of proteins with three to nine predicted transmembrane domains.
Peptide mapping of the integral membrane protein PMA1. A two-dimensional representation of PMA1 is displayed. Cylinders represent the predicted Tm domains as reported by MIPS (ref. 24). The protein segments between predicted Tm domains are drawn to approximate scale. Black lines and green cylinders represent segments of the protein not identified in this study. Red lines and the red cylinder represent segments of the protein identified in this study. One peptide was detected and identified between Tm domains 2 and 3, 10 peptides were detected and identified between Tm domains 4 and 5, and one peptide was detected and identified in the C terminus. We also detected and identified a peptide corresponding to Tm domain 5 in our analysis. The 320-amino acid domain between Tm domains 4 and 5 is the largest in the protein.
